An Agent-Based Model of Hierarchical Information-Sharing Organizations in Asynchronous Environments

Most organizations use command hierarchies—the type of hierarchy depicted in a common organizational chart—but it is not well understood why and how environments make this structure useful. One possibility is that command hierarchies provide positive net benefits when groups of agents must respond to changes in the environment, particularly when each group member’s local conditions are similar and somewhat synchronous. We ask: How does the performance of hierarchical groups vary with changing environments? We build an agent-based model to better understand the strengths and weaknesses of hierarchy for groups faced with these changes in space and time. In these environments, a local worker has more information about local conditions, but a manager has more information about overall conditions. We show that command hierarchy outperforms non-hierarchy in many synchronous and asynchronous environments, including those where local conditions differ substantially and would seem to make a manager’s “big picture†input much less useful to workers. In these more asynchronous environments, a manager’s view of overall conditions does give useful information to workers, with crucial caveats: workers must have the autonomy to judge the accuracy and relevance of manager input to their local work, or they perform worse than non-hierarchical groups. This autonomy enables the organization to learn. Relatedly, we also find increased agent memory is important for performance in all environments. Our model reveals that environments that vary locally can cause unavoidable tension between the views of front-line workers and managers, or local offices and head offices; even perfect agents find themselves in an inevitable computational dilemma. The best organizational strategy to manage this dilemma is continuing to provide manager input while enabling some degree of worker autonomy.